COVID-19
Intro
I liked the Financial Times plots for tracking the evolution of COVID-19 (https://www.ft.com/coronavirus-latest), but then they changed to different plots. So here I am more-or-less reproducing those plots (and adding some others). This is generated from an Rmarkdown document that I’ll be rerendering daily.
Country-level data
For now using the country data from https://covid.ourworldindata.org:
cases = read.csv("https://covid.ourworldindata.org/data/ecdc/total_cases.csv",
stringsAsFactors=FALSE)
cases$date = as.POSIXct(cases$date)
cases$doy = round(as.numeric(difftime(cases$date, as.POSIXct("2019-12-31"), units="days")))
deaths = read.csv("https://covid.ourworldindata.org/data/ecdc/total_deaths.csv",
stringsAsFactors=FALSE)
deaths$date = as.POSIXct(deaths$date)
deaths$doy = round(as.numeric(difftime(deaths$date, as.POSIXct("2019-12-31"), units="days")))Country-level cumulative deaths and cases
Cumulative deaths
Log-scale
Gray dashed lines are doubling times of 1, 2, 3, and 7 days (from steepest to shallowest)
Linear-scale
Cumulative cases
Log-scale
Gray dashed lines are doubling times of 1, 2, 3, and 7 days (from steepest to shallowest)
Linear-scale
Country-level incremental (daily) deaths and cases
The raw daily data is quite noisy, so there are also smoothed versions.
Daily deaths
Smoothed
Log-scale
Linear scale
Raw
Log-scale
Linear-scale
Daily new cases
Smoothed
Log-scale
Linear scale
Raw
Log-scale
Linear-scale
Country-level naive case fatality rates
By “naive” I mean: at any point in time, divide the total cumulative number of deaths by the total cumulative number of confirmed cases. This will be biased high because the denominator is too small because not all cases are detected (due to lack of testing), but on the other hand will be biased low because some active cases will result in deaths eventually. Should eventually converge on the true CFR if testing becomes widespread.
Log-scale
Linear-scale
US state-level data
So far what I’ve found is from the NY Times. It is pretty strangely structured, but oh well:
states = read.csv("https://raw.githubusercontent.com/nytimes/covid-19-data/master/us-states.csv",
stringsAsFactors=FALSE)
states$date = as.POSIXct(states$date)
states$doy = round(as.numeric(difftime(states$date, as.POSIXct("2019-12-31"), units="days")))
# Get this into a more sensible structure:
us.deaths = t(tapply(states$deaths,
list(factor(states$state, levels=sort(unique(states$state))),
factor(states$doy, levels=min(states$doy):max(states$doy))),
identity))
us.deaths = as.data.frame(us.deaths)
dates = sort(unique(states$date))
us.deaths$date = dates
doys = sort(unique(states$doy))
us.deaths$doy = doys
us.cases = t(tapply(states$cases,
list(factor(states$state, levels=sort(unique(states$state))),
factor(states$doy, levels=min(states$doy):max(states$doy))),
identity))
us.cases = as.data.frame(us.cases)
us.cases$date = dates
us.cases$doy = doysUS state-level cumulative deaths and cases
Cumulative deaths
Log-scale
Gray dashed lines are doubling times of 1, 2, 3, and 7 days (from steepest to shallowest)
Linear-scale
Cumulative cases
Log-scale
Gray dashed lines are doubling times of 1, 2, 3, and 7 days (from steepest to shallowest)
Linear-scale
US state-level incremental (daily) deaths and cases
The raw daily data is quite noisy, so there are also smoothed versions.
Daily deaths
Smoothed
Log-scale
Linear scale
Raw
Log-scale
Linear-scale
Daily new cases
Smoothed
Log-scale
Linear scale
Raw
Log-scale
Linear-scale
